JeffJournal

Learned (i.e., finally implemented) a few things today...

• As part of a continuning discussion with the guy responsible for the LAM RPM in RedHat, he suggested that I move some of the LAM data files from $(pkgdatadir) to $(sysconfdir). I resisted for a while, because $(pkgdatadir) resolves to ${prefix}/share/lam, which is nicely its own directory. $(sysconfdir) resolves to ${prefix}/etc, which means that we may have some name conflicts with files that are already in that dir. Harumph.

  We discussed/argued this for a little while, but then he played the Linux FHS card, with which I really couldn't argue. So I bit the bullet, renamed our config files to have a "lam-" prefix to guarantee a lack of name conflicts, and changed all the Makefile.am's to use sysconf_DATA instead of pkgdata_DATA. Bummer.

• But here's the interesting part: one of his points was that he wanted to be able to do "make sysconfdir=/foo/bar allÿÿÿÿ" and change various directories at "make" time, not just at ./configure time. This was problematic because I was using AC_DEFINE to set these things in the central lam_config.h at ./configure time.

  My reasoning for this was that this clears up the compile command line a lot -- no additional -D flags are necessary, and various tranlational units can just use the symbolic constants that are in lam_config.h. I still think that I'm right. :-)

  However, there is strength in being able to change such directories at compile time, especially since such behavior is documented in the autoconf docs (arf!). So I removed the relevant AC_DEFINE's from configure.in and acconfig.h and added the appropriate -D switches to all the relevant Makefile.am's.
  

(Typing this, I noticed that I tend to type an ">" automatically after ".h". This is problematic when you're composing in ÿÿÿÿHTML... does this say something Freudian about my mother?)

So I think that I realized this a few days ago when I had my last e-mail with the RedHat guy, but didn't want to admit it ('cause damnit, I'm still right!). But you really can't argue with the Linux FHS, 'cause everyone uses it...

Additionally, I guess this will influence how I hard-code all directory names in future code. <sigh>